
1
1
United States
1 year of experience
Building AI systems out of Salerno, Italy. Twelve months in, deep on Claude, the Anthropic API, and Vercel as the deployment layer. Backstory: moved to Paris at 17 to shoot fashion. Five years on Vogue and Fashion Week sets, burned out by 22, left for Portugal, did crypto and agency work before landing here. The photography years taught me workflow design under deadline pressure — which turned out to be the same skill as architecting AI systems. What I've shipped over the last twelve months, grouped roughly: Marketing systems — a content engine that turns one raw idea into a week of platform-native posts, a confidence tool for people who freeze on camera, a swipe-file system for organizing reference content. Aviation — a deal-flow system for jet charter that scrapes, normalizes, and ranks aircraft listings. Parked as a portfolio piece. Photo and media workflow — a sorter that handles the kind of cull-and-tag work that used to eat my Sundays as a photographer. Internal tools — a content hub, a personal CRM layer, half a dozen smaller utilities I use to run my own business. Most of it lives on Vercel. Some of it became products. Some became case studies. All of it taught me what actually breaks when you ship AI into real workflows.

Deals Machine is an autonomous AI sales agent that runs your entire pipeline, from first-touch outreach to closing follow-ups, while the human does the talking. Built for solo founders and small revenue teams who lose deals to follow-up debt, not to product quality. The problem. Sales teams spend the majority of their time on operational toil: logging calls, writing follow-ups, updating CRM fields, prioritizing leads. The actual selling, the 1:1 human conversation, gets squeezed into the gaps. Deals Machine flips that ratio. The solution. Deals Machine ingests sales call audio via Speechmatics, transcribes and analyzes it, synthesizes per-lead memory that the agent re-reads before every action, and then autonomously plans next steps: schedules follow-ups, drafts personalized emails, scores the lead, queues call-backs, and updates the activity log. The human just shows up and talks. Multimodal intelligence. Beyond audio, the agent reads inbound documents (RFPs, contracts, technical briefs) and screenshots to inform its reasoning. Agentic workflows. Each lead has its own agent loop with memory, tool access (CRM, calendar, email, knowledge base), and decision authority over what gets sent. Intelligent reasoning. When a lead doesn't reply or the deal stalls, the agent re-plans without human intervention. Enterprise utility. Built on Next.js + Supabase, deployed on Vercel (UI) and Vultr Ubuntu 24.04 (agent worker), with Veea Lobster Trap prompt-injection middleware hardening the reply-processing loop against malicious inbound content. Submitted solo by Kyle Dow (View 1 Studio / MAIKER) from Battipaglia, Italy.
19 May 2026